Analyzing Digital Jitter and its Components

Transcription

1 2004 High-Speed Digital Design Seminar Presentation 4 Analyzing Digital Jitter and its Components

2

3 Analyzing Digital Jitter and its Components Copyright 2004 Agilent Technologies, Inc. Agenda Jitter Overview Measurement Methodologies Why Separate Jitter? Separating Jitter with the DCA-J Page 2 4-1

4 Jitter ITU-T Definition: A Measure of the short term time variations of the significant instances of a digital signal from their ideal positions in time. MJSQ Definition: The deviation of a signal edge time at a defined amplitude of the signal from a reference time. Single transition Ideal receiver sample point The EYE Diagram Overlaid transitions Unit Interval Total Jitter Page 3 Classes of Jitter Measurements Jitter Generation How much jitter is the device under test (DUT) generating? Transmitter measurement Jitter Transfer How much of the jitter received by a DUT is passed along to the transmitted signal via the recovered clock? Repeater measurement Typically associated with synchronous networks Jitter Tolerance How much jitter on the input signal can the DUT tolerate and still produce an acceptable bit error ratio (BER) Receiver measurement Page 4 4-2

5 Jitter Methodologies & Market Segment Use WAN SAN LAN Back Plane PC & Server <= 1 Gb/s 1.5 Gb/s 3 Gb/s 5/6 Gb/s SONET/ SDH Jitter Generation Jitter Transfer Enterprise Jitter Generation requiring separation of jitter into subcomponents Proprietary Some use of Enterprise methods Digital Design Jitter Gen Cycle-to-Cycle N-Cycle 10 Gb/s Jitter Tolerance Banded Jitter Tolerance defined by stressed eye Enterprise Methodology Emerging OIF standard - CEI Page 5 Why Separate Jitter? Mechanism to enable fast estimates of Total Jitter (TJ) at low BER Tool to support budgeting of jitter in new and evolving designs Diagnostic tool for troubleshooting jitter Data Dependent Jitter (DDJ) Deterministic Jitter (DJ) Total Jitter (TJ) Random Jitter (RJ) Periodic Jitter PJ Inter-symbol Interference (ISI) Duty Cycle Distortion (DCD) Page 6 4-3

6 Enterprise jitter methodology Assess jitter in terms of its impact on system performance Bottom line is bit-error-ratio (BER) Which waveform is most likely to cause poor BER due to jitter? Let s figure out what s going on Page 7 Assessing the impact of jitter on BER System BER less than 1 error per trillion bits (10-12 ) Total jitter needs to be determined to similar probabilities There are things the eye diagram simply cannot tell us Breaking apart the jitter into its various components will efficiently lead us to an accurate measurement Page 8 4-4

7 Pruning the Enterprise Jitter Family Tree Signal jitter can be composed of several types from several mechanisms Data-Correlated Total Jitter (TJ) Data-Uncorrelated Deterministic Jitter (DJ) Random Jitter (RJ) Data Dependent Jitter (DDJ) Inter-symbol Interference (ISI) Duty Cycle Distortion (DCD) Periodic Jitter PJ Page 9 What Does Data Dependent Jitter Look Like? 270K Samples 1.35M Samples Page

8 What Does Random Jitter Look Like? 270K Samples 1.35M Samples Page 11 What Does Periodic Jitter Look Like? 270K Samples 1.35M Samples Page

9 Correlated or uncorrelated to the data? Use averaging to remove uncorrelated jitter Dominant DDJ Dominant RJ 64 avgs 64 avgs Page 13 What Does it Look Like All Together? 270K Samples 1.35M Samples Page

10 Putting together the pieces The key to understanding the complete jitter picture requires a systematic and efficient approach to isolating and quantifying the various components of jitter Just as important as pulling things apart is knowing how to put them back together for the bottom line impact on BER Page 15 Looking at Jitter Differently Page

11 Looking at Jitter Differently Page 17 Using Sampling Scope for Jitter Strengths Wide bandwidth Low intrinsic jitter Multi-functionality Scalable Historical Weaknesses Long measurement time Pattern trigger required for DDJ Errors induced by timebase delay Limited control of sampling interval No automated analysis tools Limited statistical sampling Page

12 A New Measurement Architecture Eliminate Historical Weaknesses New Triggering Hardware Efficient Sampling Technique Automated Jitter Analysis 86100C Infiniium DCA-J Page 19 New Triggering Hardware Build capability into scope to derive a pattern trigger from a clock trigger Determine pattern length and count clock pulses Generates trigger synchronous with pattern repetition PatternLock Enables precise location of samples in time Focus samples anywhere in the pattern Enables precise sampling interval Precise periodic sampling Enables optimal usage of DCA timebase Minimizes error induced by timebase delay Page

13 Efficient Sampling Technique Essential to maximize utilization of samples for jitter values Determine the amplitude vs. time function of the edge Edge Model Fix sample time at ideal edge location Page 21 Sample 1 Amplitude If amplitude is at threshold - implies no jitter Amplitude > threshold - early edge Amplitude < threshold - late edge ~100X efficiency improvement Sample 2 Jitter Ideal edge position Edge Modeling Build up samples of an edge using traditional sampling Curve fit to get best model of edge Result is mathematical transfer function of amplitude to jitter Page

14 Automated Jitter Analysis Page 23 Advanced Jitter Analysis Use Counter hardware to focus samples on edges Use averaging to isolate Data-Correlated jitter Focus on individual edges to isolate Data-Uncorrelated jitter Data-Correlated jitter is Data Dependent Jitter (DDJ) Data-Uncorrelated Jitter is made up of Random Jitter (RJ) and uncorrelated Periodic Jitter (PJ) Page

15 Data Dependent Jitter (DDJ) Correlated DJ TJ Uncorr RJ Average out the uncorrelated jitter Isolates data-correlated contributions only ISI DDJ DCD PJ Measure mean position of every edge in pattern Ideal edge position defined mean of means Obtain DDJ vs. Bit record of edge positions DDJ for a given edge is the difference between its average position and the ideal Peak-to-peak DDJ is difference between earliest edge and latest edge Build histograms for All edges Rising edges Falling edges Page 25 Duty Cycle Distortion (DCD) & Inter-Symbol Interference (ISI) Isolate rising edge data from falling edge data Difference of average locations is J DCD µ Falling - µ Rising Maximum of the peak-to-peak values is J ISI Max (P-to-P Falling, P-to-P Rising ) Correlated TJ DJ DDJ ISI DCD Uncorr RJ PJ µ F µ R DCD P-P F P-P R DDJ Page

16 Data-Uncorrelated Jitter Correlated DJ TJ Uncorr RJ Focus on only one edge at a time Isolates uncorrelated contributions ISI DDJ DCD PJ Counter hardware provides precise periodic sampling Build a time sampled record Late Periodic samples of uncorrelated jitter Jitter Time Early Build a histogram of uncorrelated jitter RJ, PJ Histogram Page 27 Random Jitter (RJ) Correlated DJ TJ Uncorr RJ Late ISI DDJ DCD PJ Jitter Early PJ spikes NOT used to derive PJ value Time FFT time sampled record Aliased jitter frequency spectrum Shows RJ & PJ PJ appears as spikes Remove PJ spikes from spectrum Interpolate across gaps left behind Resultant spectrum is made up of RJ Integrate noise power in resultant spectrum This is the random jitter (RJ) Page

17 Periodic Jitter (PJ) Correlated DJ TJ Uncorr RJ ISI DDJ DCD PJ Take RJ info from FFT and construct a Dual Dirac-delta model with appropriate slopes Match Dual Dirac-delta model to RJ, PJ histogram so that peak-peak widths match for 99.8% of volume Separation (offset) of two Gaussians corresponding to the match is the periodic jitter (PJ). Page 29 Deterministic Jitter (DJ) and Total Jitter (TJ) DDJ histogram (Data- RJ, PJ Correlated) and RJ, PJ histogram (Data- Uncorrelated) are measured directly Convolution of these histograms produces a histogram representing the PDF of all of the jitter present Total Jitter histogram * Correlated ISI DJ DDJ DCD TJ Uncorr RJ PJ DDJ Aggregate Jitter Page

18 Deterministic Jitter (DJ) and Total Jitter (TJ) Correlated TJ DJ DDJ ISI DCD Uncorr RJ PJ DJ Take RJ info from FFT and construct a Dual Dirac-delta model with appropriate slopes Match Dual Dirac-delta model to TJ histogram so that peak-peak widths match for 99.8% of volume Same technique as used to get PJ from RJ, PJ PDF DJ Separation (offset) of two Gaussians corresponding to the match is the deterministic jitter (DJ) Extrapolate down the resultant Dual Diracdelta model to the effective BER of interest (typically ) peak-to-peak deviation is TJ Page 31 Seeing the Complete Jitter Picture Page

19 Jitter Analysis on the DCA-J Low jitter timebase Lowest intrinsic RJ ~800fs rms standard ~200fs rms with 86107A Lowest Intrinsic (TJ) Wide bandwidth Up to 80 GHz Lowest intrinsic DDJ Supports wide range of bit rates Jitter Mode One button simplicity Scalable Supports all existing plug-ins Optical & electrical jitter measurements Multi-function platform Page 33 Summary Need to separate jitter in today's high data rate digital designs: Estimate Total Jitter (TJ) at very low BER Understand the sources of jitter Develop a Jitter budget New jitter characterization techniques provide more insight into jitter Resources: Jitter DCA-J Page

20 Glossary ITU-T International Telecommunications Union MJSQ Methods for Jitter and Signal Quality BER Bit Error Ratio TIA Time Interval Analyzer TJ Total Jitter RJ Random Jitter DJ Deterministic Jitter PJ Periodic Jitter DDJ Data Dependent Jitter DCD Duty Cycle Distortion ISI Inter-Symbol Interference PDF Probability Density Function Page